We collapse the buyer–seller transaction from a 6-month, 7-figure, analyst-driven workflow into a 6-week, software-mediated, intelligence-rich process. The moat is the only thing that cannot be commoditized: proprietary in-person intelligence from a distributed network of retired industry insiders, compounded by an outcome-data flywheel from every banker who uses us.
Anything reachable by an IP address becomes commoditized by foundation models on a 3–5 year horizon. Public-data moats decay to zero.
A distributed network of retired industry vets producing structured field-interview data at ~$130 per data-point. Cannot exist on an IP address by definition.
Not a better PitchBook. We own the full M&A workflow — source, qualify, intel, intro, diligence, valuation, match-make, close.
Anything reachable by an IP address is going to be open game in the next few years. Every public-data moat is on a three-to-five-year clock.
You're right. Every signal we use today lives behind a public URL — court records, regulatory disclosures, contracting databases, demographic and employment data. In 3–5 years, every one of these is trivial to scrape for any well-resourced competitor with foundation models in their hands. The signal stack we've shipped is necessary, not sufficient.
The honest question is: what does a durable moat look like in a world where every public signal is commoditized, every foundation model is replicable, and the marginal cost of building "a slightly better PitchBook" trends to zero?
Proprietary data that does not exist on any IP address — because it has not been collected yet, by anyone. In-person interview data captured by humans embedded in the verticals we serve. Each conversation is a labeled data point a competitor literally cannot acquire after the fact.
Outcome labels from every banker who uses us. They tell us which prospects closed, at what multiple, with whom, when. After a year, our model is calibrated on data nobody can buy. After three years, the gap is permanent.
A banking VP describes the kind of deal they want to find. We turn that thesis into a ranked, interviewed, warm-intro-ready pipeline. The system is two layers — statistical enrichment and human enrichment — converging into a single combined score and a warm handoff. Everything else falls out of this.
PitchBook + state SOS registrations + D&B + LinkedIn + our own scraped private-company corpus, cross-referenced for every company that fits the VP's filtration spec.
We run every candidate through three converging signal streams:
For the top-statistical-score prospects, we send a homie:
The statistical score (Layer 1) and the interview-derived features (Layer 2) feed into a single calibrated conviction score. For each prospect, the banker receives:
The operative has already met the owner — they make the warm intro. Netra delivers the intro-email template (referencing specifics from the interview), calendar coordination, and a deal-context briefing. No cold outreach. No "did I catch you at a bad time?" The banker meets a vetted, conviction-scored prospect whose context they already know.
A distributed network of retired industry vets — former operators, GMs, founders — paid $50/hr to drive to target companies and conduct structured 30-minute interviews. Captures: succession willingness, recent capex decisions, customer concentration anecdotes, owner mood, regulatory exposure, family dynamics.
Each interview is a structured data row competitors cannot scrape, archive, or reverse-engineer. It does not exist on any IP address.
Every banker who uses Netra tells us which prospects closed, at what valuation, with whom, when. Each label is a training point no competitor can buy from PitchBook or anywhere else.
After 12 months of paying customers: ~5,000 outcome labels. After 36 months: ~50,000. Our model calibration becomes structurally unreachable to a new entrant — they would have to wait three years and lose money the whole time to catch up.
Bankers live in Affinity, HubSpot, Salesforce. Netra writes ranked scores, intelligence summaries, and intro-pathways back into their CRM as native fields. Within 6 months a banker's morning starts inside Netra. Within 12 months, ripping it out means rebuilding their daily process.
Switching cost climbs from zero to substantial — the same dynamic that made Bloomberg Terminal unkillable for 40 years.
Investment banking is fundamentally a trust business. The firms most likely to win mandates have decades of relationship capital. Becoming the brand for LMM signal intelligence means every banker introducing a deal references Netra by name, and every PE bizdev team checks Netra before they cold-call.
Network effects + reputational compounding are the slow-burning moat that decacorn companies are built on. SourceScrub, Grata, Affinity each occupy a piece of this — Netra owns the full picture.
Sourcing, qualifying, intel, intro — every stage where data, AI, and in-person intelligence compress the work that determines whether a deal happens. From the moment a VP describes a thesis to the moment they sit down with an owner who's already been validated as ready to transact, Netra owns the pipeline end-to-end.
PitchBook keyword filters → 5,000 names → analyst trims to 500. Netra ranks at AUC 0.831 on the 1,369-row mfg holdout.
Top-decile lift 2.55× on mfg cuts banker call volume by ~2.5×. Same hit-rate at one-third the effort.
Field-operative network produces structured interview data on top-ranked prospects. The moat lives here.
Operative who interviewed the owner brokers the intro. Banker walks in pre-qualified, pre-briefed, pre-warm.
Modeled on a typical LMM banking VP working a 100-prospect cohort in manufacturing. Today's process is reconstructed from interviews with practicing LMM bankers. The Netra column reflects v1.0 capabilities — field network active, full workflow shipped.
The second-order consequence: a banking VP can run their own deals without a 3-person analyst pod. That is what Jeffrey's "kill the analyst layer" thesis actually means in practice. We are not replacing analysts; we are giving VPs the ability to operate the way only senior partners do today.
Every customer is also a data collector for us. A banker who works a 200-row prospect list and reports back "these 12 closed, these 188 did not" hands us 200 labeled training points. Nobody else can buy those labels — they live inside that banker's CRM.
After 5 paying customers running ~200 prospects/month each, we ingest ~12,000 labeled outcomes per year. After 20 customers: 48,000/year. The model recalibrates monthly on data that compounds with our customer count, not with our engineering hires.
Combined with the field-operative intelligence layer, every cycle widens the calibration gap between us and any new entrant — including a YC team in 2028 with Claude Code who can rebuild the v0.8.5 signal stack in a weekend but cannot replay three years of outcome data.
The most common objection to a human-in-the-loop data product is that it doesn't scale. We modeled it carefully — and at the scale we need to reach decacorn outcomes, the field network is both the moat AND the most efficient unit of customer value creation.
For context: a single Sourcescrub or Grata seat costs an LMM bank $15K–$50K/yr and provides commodity public data. Our interview-validated signal stack is materially more useful per dollar — and the data we generate accumulates as a permanent asset, not a recurring expense.
The field network operates at a loss as a customer-cost line item, but it is funded as a moat investment the way pharma funds clinical trials — the resulting dataset is the company's most valuable asset. By year 4 we monetize the dataset directly via acquirer-side access fees, and the field unit goes EBITDA-positive on its own.
The current product, in isolation, is a sourcing tool — a feature, with a ceiling. The reframed product is the origination and intelligence layer for private-market M&A, with a moat that compounds and a data asset that monetizes twice. Two revenue lines, both anchored in real customer willingness-to-pay today.
Banker-side SaaS. VPs, directors, and analysts at LMM advisory banks. PitchBook charges these seats $30K+/yr. Affinity sits at $5K+/seat/yr. Capital IQ and Sourcescrub price in similar bands. Netra prices into the same line item with a meaningfully better product at the front of the deal funnel.
Acquirer-side data licensing. PE bizdev teams, family offices, search funds, strategic acquirers. The dataset our field network produces is exactly the input these firms pay $50K+/yr for in PitchBook Insights and equivalents — except ours has the in-person intelligence layer they can't get anywhere else.
The architectural choice — own the origination workflow and the proprietary dataset that backs it — supports a meaningfully higher ceiling than a pure SaaS competitor at the same price point, because the data asset is independently monetizable on the buy-side. The marketplace / transaction-fee layer is a longer-term option we're happy to walk through in conversation, but it isn't load-bearing for this thesis.
Public data is the floor.
In-person intelligence is the moat.
Workflow ownership is the gold mine.